Data Scientist, Forecasting and Optimisation

About the role

This is a full time, on site role in Bengaluru. You will design and build the models that decide what stock moves, working with product, engineering and customer-facing teams. Models go from problem definition through prototype to production Rust.

Key responsibilities

• Forecast demand at SKU, store and week level, where most series are sparse and intermittent, including cold start on new product with no sales history.
• Build hierarchical reconciliation so store forecasts sum coherently to region and national plans, along with size and colour curves, cannibalisation and halo across a range, and seasonality that follows the retail calendar rather than the Gregorian one.
• Build the optimisation models behind allocation and replenishment, safety stock policy, assortment and range planning, and markdown timing and terminal stock risk.
• Build the simulation layer behind our scenario tools, so that when a merchandiser asks what happens if they shift ten per cent of stock online, something answers credibly and fast.
• Work on the reasoning layer behind our autonomous workflows: retrieval over retail context, tool use, natural-language explanation of forecasts, and the evaluation harnesses that tell us whether any of it works.
• Measure causal effects: promotional lift, price elasticity, and what would have sold without the discount.
• Set the standards for rolling-origin backtesting, honest baselines, error metrics that reflect commercial cost, data quality checks, model documentation and auditability. Our outputs write back into customer ERPs, so a retailer has to be able to trust an automated purchase order.

What we are looking for

• Three to five years building and deploying machine learning in production. What you shipped matters more than your title.
• Depth in at least two of time series forecasting, mathematical optimisation, simulation, or causal inference. All four is rare and we will not pretend otherwise.
• Genuine software engineering ability rather than scripting. Tested, reviewed, maintainable code, comfort with static types, and enough understanding of ownership, allocation and concurrency that Rust is new syntax rather than a new concept.
• Evidence you have learned something hard and unfamiliar fast and shipped with it. Tell us that story in your application, it is the thing we will dig into.
• Strong SQL. Window functions and query plans, not just joins.
• Experience with messy operational data from enterprise source systems: backdated corrections, missing weeks, hierarchies that change mid-year, schemas nobody documented. If you have reconciled data across SAP, NetSuite or Dynamics, say so.
• Clear communication with non-technical stakeholders. You will explain a forecast to a Head of Buying placing serious money into stock.
• Experience with LLM systems in production, particularly evaluation: eval sets, regression measurement, cost and latency management, and guardrails.